CHOOSEBusiness workflows
Jev for Transaction Classification
Jev is a tool that classifies financial transactions into categories, demonstrating significant speed and cost improvements over existing models for this task.
Pattern✦ model performance benchmarks
Jev is seriously impressive and it's going to be a total game-changer for enterprise automation workflows.
Finally got through the waiting list this morning (thank you @typesafeai) and started testing it in a few of my personal projects.
Here's a real example that shows its utility. I have a personal finance app where I upload my credit card and bank statements, and pass each transaction line to an LLM to classify them into categories (shopping, dining, groceries etc...). Thought that would be the perfect test for Jev and it is seriously fast. For this specific task, it turned out to be 3x faster and anywhere from 11x to 38x cheaper than this leading lab's models.
This is going to be a huge deal for enterprise automation use cases.
PS: I have only checked the speed and cost for this run. I haven't thoroughly tested accuracy yet. For instance, my "Challenger PLQ" spend left uncategorized by Jev, but an LLM easily tagged it as shopping.
I’ll be running more tests soon. Let me know if you want to hear about the results, or drop me a DM if you want to figure out how to work this into your pipelines.
